Ventricular shape of monozygotic twins discordant for schizophrenia reflects vulnerability

1
Subjects and image data

• Subjects Image data provided by D. Weinberger,
  NIMH, Bethesda 2
• MZ 10 healthy monozygotic twin pairs (N210)
• DS 9 MZ twin pairs discordant for
  schizophrenia (N29)
• DZ 10 dizygotic twin pairs, all healthy controls
  (N210)
• NR Selection of unrelated, healthy subject pairs
  with best possible match of age and gender
  (N210)
• Image data Gradient-echo T1w (256x256x128,
  240mm FOV, 1.5mm slice distance)

Ventricular shape of monozygotic twins discordant
for schizophrenia reflects vulnerability 2M
Styner, 1,2G Gerig, 3DW Jones, 3DR Weinberger,
1JA Lieberman Dept. of 1Psychiatry and
2Computer Science University of North Carolina,
Chapel Hill, NC 27614, USA 3 National Institute
of Health, Bethesda, MD gerig_at_cs.unc.edu /
http//www.cs.unc.edu/gerig
METHODS
The T1w high resolution MR image data are
processed by automatic brain tissue segmentation
followed by shape parameterization.

• Automatic, atlas-based 3D voxel segmentation
  technique.
• Segmentation of lateral ventricles by manually
  guided 3D connectivity.

Summary
Twin B

• Enlarged ventricular size and/or asymmetry have
  been found markers for psychiatric illness,
  including schizophrenia. We studied ventricular
  size and shape in volumetric MRI of dizygotic
  normal twin pairs (DZ), monozygotic normal twin
  pairs (MZ), and monozygotic twin pairs discordant
  for schizophrenia (DS), subdivided into affected
  (DSS) and non-affected (DSH). A fourth group of
  unrelated matched pairs (NR) was selected from
  the two normal groups.
• Left and right ventricles were segmented from
  high resolution T1 SPGR MRI followed by
  surface-based shape parametrization. Shape
  differences between pairs of shape was measured
  as mean average difference between corresponding
  surface points.
• The statistical analysis included two tests with
  corrections for age and gender
• First, we investigated pairwise shape similarity
  between ventricles of co-twins. The group
  difference of co-twin similarity between normal
  MZ and discordant DS was not significant, whereas
  both groups were significantly different from NR.
  The pairwise co-twin shape similarity seems
  equally large for healthy MZ and for MZ-DS,
  reflecting morphologic similarity due to
  heritability.
• Second, we examined the shape difference of the
  affected and unaffected DS subgroups in
  comparison to the normal control group. The
  average shape of normal co-twins not included in
  group tests served as a template for comparisons.
  Both the affected and unaffected DS groups showed
  significant shape difference from the normal
  population. These tests show ventricular shape
  alterations from healthy controls not only in the
  affected (DSS) but also in the non-affected
  groups (DSH). This leads to the conclusion that
  ventricular shape change might reflect
  vulnerability for schizophrenia and hence be a
  marker for a neurodevelopmental aspect of the
  illness.
• Both statistical tests applied to volumes did
  not show any differences between MZ and DS
  groups, suggesting that shape analysis is more
  sensitive to subtle structural changes than
  volumetry.

Twin A

• Parametrization of object surfaces using 3D
  Fourier harmonics (SPHARM. 1,3).
• Spatial alignment of structures by Procrustes fit
  of sets of homologous surface points.
• Size normalization by individual volumes.
• Pairwise shape difference Signed and unsigned
  mean average differences (MAD).

MZ
Twin B
Twin A
DZ
Hierarchical surface parameterization by SPHARM
Parameterized object surfaces showing
correspondences.

• Shape Distance Metric
• Left Figures
• Two lateral ventricles showed after alignment.
• Same as A, mesh overlay.
• Surface with color-coded shape distances.
• Right Figures
• Statistical shape analysis providing a
  significance map (blue non-significant, red
  highly significant)

March 2003 2
March 2003 1
2
RESULTS
Group differences between ventricle shapes
Volume Analysis
Lateral ventricle shapes after volume
normalization. Ventricles of 10 DS twin pairs
(left) and 5 MZ and 5 DZ pairs (right) are shown
side by side.
Volumes corrected for ICV, age and gender.
Affected twins (DSS) and nonaffected twins (DSH)
do not differ from healthy controls.
Pairwise Co-Twin Differences
Co-twin Volume Difference
Shape distance to healthy control template
Shape differences to healthy control template,
normalized for unit size and corrected for gender
and age. Ventricular shapes of affected twins
(DSS) differ significantly from the healthy
controls (L plt0.039, R 0.058). Ventricular
shapes of non-affected twins (DSH) show even more
significant differences from healthy controls (L
plt0.0042, R plt0.0089). Affected versus
non-affected twins show no significant group
difference.
Absolute volume differences between co-twins.
Volumes were corrected for ICV. Statistics of
pairwise differences is corrected for age and
gender. There is a trend MZ MS DZ lt NR, but
group tests were not significant except DS versus
NR and DZ versus NR.
Co-twin Shape Similarity
Shape differences between co-twins after
normalizing for size differences. There is a
significant decrease of shape similarity with
decreasing genetic similarity MZ DS lt DZ lt
NR. The pairwise shape differences between
healthy MZ and MZ-DS are not significantly
different.
Pointwise averaged distance maps to healthy
control template for each group. Distances are
color-coded to show localization of shape
differences.
CONCLUSIONS

• Analysis of ventricle volumes and volume
  differences Not significant due to large
  variability in all groups.
• Pairwise co-twin shape differences after volume
  normalization MZ ? DS lt
  DZ lt NR, reflecting morphologic similarity due to
  heritability.
• Shape differences between groups We found
  ventricular shape alterations in affected (DSS)
  and non-affected (DSH) twins. This might lead to
  the conclusion that ventricular shape might
  reflect vulnerability for schizophrenia and might
  be a marker for neurodevelopmental aspects of
  illness.

Visualization of the pointwise co-twin distance
maps for each group (residuals after correction
for gender and age, group averages). The
distances are color-coded to show absolute
differences between 2 and 8mm. The figures
illustrate the decreasing shape similarity MZ
DS lt DZ lt NR. Healthy MZ are not significantly
different from MZ discordant for schizophrenia
(DS).
References 1 A. Kelemen, G. Székely, and G.
Gerig, Three-dimensional Model-based
Segmentation, IEEE Transactions on Medical
Imaging (IEEE TMI), 18(10)828-839, Oct 1999 2
A. Bartley, D. Jones, and D. Weinberger, Genetic
variability of human brain size and cortical
patterns,Brain, vol. 120, pp. 257269, 1997. 3
G. Gerig, M. Styner, D. Jones, D. Weinberger, and
J. Lieberman, Shape Analysis of brain ventricles
using SPHARM, in Proc. Workshop on Math.
Methods in Biomed. Image Analysis MMBIA 2001,
IEEE Comp Soc, pp. 171-178, Dec. 2001
March 2003 4
March 2003 3